Keras Image Helper

A lightweight library for pre-processing images for pre-trained keras models

Imagine you have a Keras model. To use it, you need to apply a certain pre-processing function to all the images. Something like that:

from tensorflow.keras.applications.xception import preprocess_input

What if you want to now deploy this model to AWS Lambda? Or deploy your model with TF-Serving? You don't want to use the entire TensorFlow package just for that.

The solution is simple - use keras_image_helper

Usage

For an xception model:

from keras_image_helper import create_preprocessor

preprocessor = create_preprocessor('xception', target_size=(299, 299))

url = 'http://bit.ly/mlbookcamp-pants'
X = preprocessor.from_url(url)

It's also possible to provide custom functions instead of preprocessor name:

def custom_function(x):
    x = x / 255.0
    x = x.round(1)
    x.transpose(0, 3, 1, 2)
    return x

preprocessor = create_preprocessor(custom_function, target_size=(3, 3))

Now you can use X for your model:

preds = model.predict(X)

That's all :tada:

For more examples, check test.ipynb

Currently you can use the following pre-processors:

• xception
• resnet50
• vgg16
• inception_v3

If something you need is missing, PRs are welcome

Installation

It's available on PyPI, so you can install it with pip:

pip install keras_image_helper

You can also install the latest version from this repo:

git clone git@github.com:alexeygrigorev/keras-image-helper.git
python setup.py install

Publishing

1. Install development dependencies:

uv sync --dev

2. Build the package:

uv run hatch build

3. Publish to test PyPI:

uv run hatch publish --repo test

4. Publish to PyPI:

uv run hatch publish

5. Clean up:

rm -r dist/

Note: For Hatch publishing, you'll need to configure your PyPI credentials in ~/.pypirc or use environment variables.

PyPI Credentials Setup

Create a .pypirc file in your home directory with your PyPI credentials:

[distutils]
index-servers =
    pypi
    testpypi

[pypi]
repository = https://upload.pypi.org/legacy/
username = __token__
password = pypi-your-main-api-token-here

[testpypi]
repository = https://test.pypi.org/legacy/
username = __token__
password = pypi-your-test-api-token-here

Done!